Optimizing dialysis water treatment based on medical planning requirements

Abstract

This paper introduces a new approach to improving hemodialysis and peritoneal dialysis processes. This maintains compatibility between the water treatment unit and active dialysis machines. Identification of different deviations in the patient’s requirements to classify the suitable flow rate (FR) of the pure water can be created through an Adaptive Neuro-Fuzzy Inference System (ANFIS). Four types of categories are classified: quarter load, half load, three-quarter load, and full load. This includes two main steps: identifying the dialysis categories with patients and measuring the suitable FR for each category. The effective FR of water was computed to indicate how the number of pumps rose as the number of patients treated within the dialysis unit increased. The introduced prediction method was tested against artificial neural networks to predict alum doses (ANN-AD) method, artificial intelligence as a combinatorial optimization strategy (AN-COS) using peach-palm waste in a solid-state fermentation environment, expert system applications (ESA) in chronic kidney disease management (MDSSs), prescription optimization for automated peritoneal dialysis (BCPH). In this work, pumping system parameter estimation was utilized to assess the performance of the introduced systems based on input voltage, input current, rotor speed, and electromagnetic torque waveforms of the pumping drive. This method uses the ANFIS algorithm to construct a high-performance pumping system to evaluate and predict the flow rate of dialysis machines based on input–output estimates. The flow rate values for each patient in critical care units can be controlled by the presented algorithm according to the time of dialysis. A good predictive model based on integrating the experience and decision making of medical planners was implemented. The proposed method not only improves accuracy but also saves time in the dialysis unit. The presented fuzzy classifier uses the ratio of the number of predicted true positives and false positives to the total number of anticipated positives. The simulation results show that the proposed technique can be successfully used to classify the suitable FR of the water according to the hemodialysis load. One of the advantages of the presented technique is the ability to predict the value of FR in real time with reliability, low computational complexity, and a high accuracy of 98.01%, which simplifies the dialysis process for many patients while at the same time leading to improved performance of hemodialysis. The presented method can be implanted in online FR estimation and analysis hardware for hemodialysis, is easy to calculate and run in real time, minimizes hemodialysis costs, and saves run time.